OUP user menu

Syncope in the emergency department: comparison of standardized admission criteria with clinical practice

Marcos Daccarett, Tawni L. Jetter, Stephen L. Wasmund, Michele Brignole, Mohamed H. Hamdan
DOI: http://dx.doi.org/10.1093/europace/eur201 1632-1638 First published online: 14 July 2011

Abstract

Aims Syncope is a major health care problem that accounts for many emergency department (ED) and hospital admissions. This study was conducted to investigate the short-term risk of serious events in patients presenting to the ED with syncope and to compare guideline-based admission criteria with those adopted in clinical practice.

Methods and results A single-centre retrospective analysis was performed on ED visits between January and June 2009. We used the ICD-9 code 780.2 for syncope as the primary diagnosis. The prevalence of serious events within 7 days of the index presentation was evaluated. In addition, admissions and discharges were classified as being appropriate or inappropriate based on standardized guideline-based criteria integrated in a new Faint-Algorithm developed at the University of Utah. Two hundred and fifty-four ED visits met the inclusion criteria. One hundred and thirty-six patients were discharged home and the remaining 118 were admitted. The prevalence of serious events in the discharged and admission groups were 5 and 10, respectively, (P= NS). According to the Faint-Algorithm, the number of inappropriate discharges and admissions were 8 out of 136 and 69 out of 118, respectively. Using the Faint-Algorithm, only 57 patients instead of 118 patients should have been admitted resulting in a 52% reduction in admission rate. Furthermore, in the remaining 197 patients who should have been discharged, the prevalence of serious events was not significantly different than that observed in the 136 patients who were actually discharged (3% vs. 4%).

Conclusion There are significant numbers of inappropriate discharges and admissions in patients presenting with syncope. The standardized guideline-based criteria integrated in the new Faint-Algorithm provide promise but require further prospective evaluation.

  • Syncope
  • Algorithm
  • Emergency department

Introduction

Syncope is a major health-care problem with a yearly cost exceeding $2 billion in the USA.14 It accounts for nearly 1.5% of all emergency department (ED) visits and up to 6% of hospital admissions.58 While the majority of syncope evaluation is performed in the outpatient setting, the overall cost is primarily driven by inpatient admissions.3 Several studies have shown that many admissions for syncope are unwarranted.6,9 Despite the presence of multiple specialty guidelines, practice in the ED remains conservative with most physicians erring on the side of caution particularly with the elderly patient. While the intent of such practice is to provide the best care, it has never been shown to improve outcome or reduce mortality in the majority of patients. Therefore, the challenge remains in risk stratification at the time of presentation.

Rules and risk stratification schemes have been developed to help with admission decisions;1013 however, it is unclear as to whether these have led to a change in practice.1417 While many studies have shown promise in predicting future arrhythmias or cardiac syncope, the methodology was often based on long-term follow-up. It is hard to justify short-term management strategies and the need for emergency hospitalization based on long-term outcome such as 1-year follow-up. Several studies have assessed the short-term risk in patients presenting to the ED.11,12,1820 In general, they showed that only a small percentage of patients admitted experienced a life-threatening event during the subsequent 7–30 days.

Since the risk of life-threatening events in the few days or weeks following the index event is the main reason for admission, we sought to assess the short-term risk of life-threatening events in patients presenting to the ED with syncope. Moreover, we compared a new set of admission criteria derived from the most recent American21 and European22 guidelines with those adopted in clinical practice.

Methods

Study design

A single-centre retrospective analysis was performed on data collected between 1 January 2009 and 30 June 2009. This study was approved by the Institutional Review Board at the University of Utah.

We performed an electronic search of the patient database at the University of Utah to identify patients who presented to the ED with the primary diagnosis of syncope. The University Hospital is a tertiary care hospital located in Salt Lake City, UT, USA.

We used the ICD-9 code 780.2 ‘Syncope and Collapse: Blackout; Fainting; Near-Syncope; Vasovagal attacks.’ Patients who presented with a secondary diagnosis of syncope were excluded on the basis that these patients might have had obvious reasons for syncope that were coded in the primary diagnosis position such as myocardial infarction, arrhythmia, haemorrhage, or stroke. In addition, patients whose presentation was not listed as their first ED-visit during the data collection period were excluded.

Measurements

Patients were categorized based on whether they were admitted or discharged from the ED. Our primary outcome was any predefined serious clinical event that occurred during the 7-day period after the initial ED evaluation. The prevalence of serious events within 7 days of the index presentation was assessed by reviewing the electronic patient record system, hospital pacemaker records and radiological reports. Serious events included (i) acute myocardial infarction, (ii) implantation of a pacemaker or cardiac defibrillator within 1 week of the index collapse, (iii) pulmonary embolus confirmed on lung perfusion scan or CT pulmonary angiography, (iv) cerebral vascular accident, intracranial haemorrhage, or subarachnoid haemorrhage demonstrated by brain imaging or lumbar puncture, (v) haemorrhage requiring a blood transfusion of >2 units packed red blood cells, (vi) acute surgical procedure or endoscopic intervention, (vii) return ED visit or hospitalization for a related event, (viii) potentially life-threatening arrhythmia [ventricular fibrillation, sustained ventricular tachycardia, ventricular pause (>3 s), or asystole], (ix) heart failure, and (x) death. Death was confirmed by findings in the medical record, Social Security Death Index, or death certificate. One cardiologist (M.D.) and a nurse practitioner (T.L.J.) reviewed all the serious events. In case there was no consensus, a third cardiologist (M.H.H.) was consulted to resolve any conflict.

In addition, patients were further classified as either being appropriately or inappropriately admitted or discharged based on a new set of admission criteria derived from the most recent American21 and European22 guidelines. It consists of 26 admission criteria derived from history, physical exam, 12-lead electrocardiogram, trans-thoracic echocardiogram, and laboratory findings. They are classified under four categories including (i) cardiac-arrhythmic causes, (ii) cardiac-ischemic causes, (iii) cardiovascular and pulmonary structural causes, and (iv) non-cardiovascular causes. The list of the admission criteria with the above classification and the diagnostic tools used are provided in Table 1. In order to maximize the standardization of the diagnostic pathway and to avoid unwanted individual differences in practice, we implemented the above criteria into a new Faint-Algorithm developed at the University of Utah. The Faint-Algorithm is a web-based interactive algorithm, which integrates the guidelines’ recommendations for risk assessment and admission in a structured diagnostic pathway. Clinical data are entered in structured electronic forms. At the end of data entry, the algorithm states whether an admission is ‘indicated’ or ‘not indicated’. In summary, we assessed the presence of any discrepancies between clinical practice and the output from the guideline-based algorithm with respect to the need for admissions.

View this table:
Table 1

Faint-Algorithm admission criteria (adapted from 2009 ESC guidelines on Syncope)

Reason for admissionDiagnostic criteria
Cardiac-arrhythmic causes
  1. Sinus bradycardia<40 beats/min or pauses >3 s

  2. Mobitz II or 2:1 second-degree or third-degree atrioventricular block

  3. Alternating left and right bundle branch block

  4. Sustained supraventricular tachycardia

  5. Sustained ventricular tachycardia

  6. Pacemaker (ICD) malfunction with cardiac pauses.

  7. LBBB or RBBB+left/right axis deviation

  8. Long-QT pattern

  9. Brugada pattern

  10. ARVD pattern

  11. WPW pattern

12-lead standard ECG
Cardiac-Ischaemic causes
  • (12) Cardiac ischaemia

 
Chest pain and troponin abnormal
Cardiovascular and pulmonary structural causes
  • (13) Prolapsing atrial myxoma, tumour

  • (14) Severe aortic stenosis

  • (15) Respiratory insufficiently defined as shortness of breath and O2 saturation <70%

  • (16) Acute aortic dissection

  • (17) Pericardial tamponade

  • (18) Severe hypertrophic obstructive cardiomyopathy (HOCM)

  • (19) Severe prosthetic valve dysfunction

 
Echocardiogram (items 13–19 and 21)
  • (20) Sustained (≥2 measurements at >5 min) supine systolic hypotension ≤80 mmHg

 
Systolic blood pressure measurement (item 20)
  • (21) Severe systolic dysfunction (e.g. <40%)

 
  • (22) History of myocardial infarction with mild LV dysfunction (LVEF>40%) and absence of criteria for vasovagal syncope or orthostatic hypotension

 
Echocardiogram+patient's history (item 22)
Non-cardiovascular causes
  • (23) Acute haemorrhage

 
Hematocrit <30
  • (24) End-stage diseases (cancer, renal dialysis, etc.)

 
Patient's history
  • (25) Major physical injuries secondary to syncope

 
Physical examination
  • (26) Minor physical injuries and symptomatic orthostatic hypotension

 
Physical examination and orthostatic blood pressure measurement
  • ESC, European Society of Cardiology; ICD, implantable cardioverter–defibrillator; LBBB/RBBB, left/right bundle branch block; ARVD, arrhythmogenic right ventricular Dysplasia; WPW, Wolff−Parkinson−White; LVEF, left-ventricular ejection fraction.

Statistical analysis

Analysis was performed using SAS 9.2 statistical software (SAS Institute Inc., Cary, NC, USA). Continuous variables are reported as mean ± standard deviation. The Shapiro−Wilk statistic and graphic analysis of residuals was used to assess the normality and constancy of residuals. Continuous variables were evaluated using a one-way analysis of variance. Categorical values were evaluated using Fisher's exact test or χ2 test when appropriate. The evaluation of the admission and discharge agreement between the observed clinical practice and the Faint-Algorithm (inter-rater reliability) was performed using the kappa and Bowker test.

Results

Characteristics of study subjects

During the 6-month study period, there were 19550 ED visits to the University of Utah. The diagnosis of syncope was found in 329 visits (1.7%) with 254 (1.3%) meeting the inclusion criteria. Seventy-five patients were excluded for the following reasons: 40 patients were seen in an earlier visit with the same complaint, 18 patients had another diagnosis as their primary diagnosis [neurologic (n= 8), trauma (n= 3), hypoglycemia (n= 2), anemia (n= 2), substance abuse (n= 2), and cardiac arrest (n= 1)], 14 patients were found not to have loss of consciousness after reviewing their medical history, 2 patients had missing records, and 1 patient left against medical advice. A flow diagram showing patient enrollment is provided in Figure 1. A summary of the clinical characteristics of these patients is provided in Table 2.

View this table:
Table 2

Patients’ clinical characteristics

Admitted (n= 118)Discharged (n= 136)P
Age (mean ± SD)47.1 ± 21.446.9 ± 21.4NS
Female gender60 (50.1%)78 (57.3%)NS
EGSYS score (mean ± SD)0.87 ± 1.890.14 ± 1.530.01
Patient with EGSYS score≥331 (26.3%)17 (12.5%)0.005
Past medical history
History of arrhythmias10 (8.5%)10 (7.3%)NS
History of CAD6 (5.1%)11 (8.1%)NS
History of myocardial Infarction4 (3.4%)8 (5.9%)NS
History of hypertension30 (25.4%)30 (22.1%)NS
History of heart failure1 (0.8%)3 (2.2%)NS
Previous cardiac surgery5 (4.2%)10 (7.3%)NS
History of hyperlipidemia12 (10.2%)9 (6.6%)NS
History of pulmonary disease0 (0%)1 (0.7%)NS
History of diabetes mellitus17 (14.4%)15 (11%)NS
Prior neurological history10 (8.5%)9 (6.6%)NS
  • NS, not significant; EGSYS, Evaluation of Guidelines in SYncope Study; CAD, coronary artery disease.

Figure 1

A flow diagram showing patient enrollment into the validation cohort of the Faint-Algorithm.

Discharges, admissions, and serious events' rates based on current practice

Of the 254 patients who met the inclusion criteria, 136 (54%) patients were discharged home. The remaining 118 (46%) patients were admitted to either the observation unit (22%) or the in-patient unit (24%) for further monitoring and evaluation. The prevalence of serious events in the discharged and admission groups were 5 out of 136 (4%) and 10 out of 118 (8%), respectively (P= NS) (Table 4). All five serious events that occurred in the discharged group were syncopal recurrences in patients without structural heart disease. Furthermore, there were no reported complications resulting from these relapses.

Appropriateness of discharges and admissions according to the Faint-Algorithm

According to the University of Utah Faint-Algorithm, the number of inappropriate discharges and admissions were 8 out of 136 (6%) and 69 out of 118 (58%), respectively. The eight patients who were inappropriately discharged met the following admission criteria of the Faint-Algorithm: bifascicular block (n= 2), QT prolongation (n= 2; QTc = 467 and 484 ms), previous myocardial infarction in the absence of typical vasovagal like symptoms or orthostatic hypotension (n= 1), and major trauma (n= 3). When assessing the appropriateness of admissions and discharges, a high discrepancy of indications (κ = 0.35, P< 0.001) was observed.

The prevalence of serious events in the appropriately admitted group (n= 49) was significantly greater than in the appropriately discharged group (n= 128) (16 vs. 3%, respectively, P< 0.01). However, the prevalence of serious events in the inappropriately discharged group (n= 8) was not significantly different than in the appropriately admitted group (n= 49) (13 vs. 16%, respectively, P= NS). Similarly, the prevalence of serious events in the inappropriately admitted group (n= 69) was not significantly different than in the appropriately discharged group (n= 128) (3 vs. 3%, P= NS). A summary of the number of appropriate and inappropriate discharges and admissions according to the Faint-Algorithm and the prevalence of serious events in each subgroup is provided in Table 3.

View this table:
Table 3

Appropriateness of admissions, discharges and prevalence of serious events according to the Faint-Algorithm

Discharges (n= 136)Admissions (n= 118)
Appropriate (n= 128)Inappropriate (n= 8)Appropriate (n= 49)Inappropriate (n= 69)
Serious events within 7 days after visit4 (3%)1 (13%)8 (16%)2 (3%)
Acute myocardial infarction
Pacemaker or ICD1
Pulmonary embolism1
CVA or intracranial haemorrhage
Haemorrhage requiring >2U PRBC2
Acute surgical procedure
Recurrence of symptoms/readmission4111
Life threatening arrhythmias3
Congestive heart failure
Death1
  • ICD, implantable cardioverter defibrillator; CVA, cerebral vascular accident; PRBC, packed red blood cells.

Comparison between clinical practice and Faint-Algorithm

Using the Faint-Algorithm, only 57 (22%) patients instead of 118 (46%) patients should have been admitted resulting in a 52% reduction in admission rate (P< 0.001). Furthermore, in the remaining 197 patients that should have been discharged, the prevalence of serious events was not significantly different than what was observed in the 136 patients who were actually discharged (3 vs. 4%, respectively, P= NS). Based on the Faint-Algorithm, five out of six serious events in the discharged group would have been non-complicated syncopal recurrences in patients without structural heart disease. A summary of the comparison between clinical practice and Faint-Algorithm is provided in Table 4.

View this table:
Table 4

Comparison between clinical practice and Faint-Algorithm indications among 254 patients referred to the ED

Clinical practiceFaint-Algorithm
DischargesAdmissionsDischargesAdmissions
Number (%)136 (54%)118 (46%)197 (78%)57 (22%)
Serious events within 7 days after the index visit (%)5 (4%)10 (8%)6 (3%)9 (16%)
  • ED, emergency department.

Discussion

The main findings from our study include the following: First, the prevalence of serious events in the admission group was low and did not justify most of the admissions. Second, according to the newly developed software integrating the guideline-based criteria, 6% of the discharges and 58% of the admissions were inappropriate. The utilization of the guideline-based criteria would have allowed a safe 52% reduction in admission rate without a significant difference in the prevalence of serious events in the discharged group. The above results suggest that the newly developed software integrating the guideline-based criteria could result in an improvement in patient care and a reduction in cost by decreasing the number of inappropriate discharges and admissions. Prospective and external validation of the algorithm is needed to test the above hypothesis.

The need to reduce the number of inappropriate admissions

Most studies in the literature have shown that there are an excessive number of unwarranted admissions in patients presenting with syncope. Furthermore, the event rate in the first 30 days following the index event is low ranging between 6.1 and 7.3%.11,1820 Our findings are consistent with these studies showing a significant number of inappropriate admissions with an event rate of 4 and 8% in the discharged and admitted groups, respectively. It is understood that when in doubt, a physician will admit the patient. However, it is undeniable, that hospital admissions are expensive and begets further tests and consults that may prove unnecessary. In addition, inappropriate admissions are inconvenient and hazardous to patients due to bed pressures and exposure to nosocomial infections. Therefore, there is clearly a need to improve the management of patients presenting with syncope to the ED.

There are two methods of safely reducing the number of inappropriate admissions without increasing the risk of serious event in discharged patients. One method is to focus on triaging patients by means of validated score-questionnaires. Another method is to apply a structured guideline-based algorithm similar to the Faint-Algorithm used in the present study.

Risk stratification score-questionnaires

There have been several syncope risk stratification studies aimed at helping the clinician in the ED. However, most of these studies were limited by the small sample size, lack of validation, or absence of short-term risk assessment.10,12,13,18,2325 In fact, to our knowledge, only few studies listed below assessed the short-term risk in patients presenting to the ED with syncope.

In the San Francisco study,12,26 physicians prospectively completed a structured data form when evaluating patients presenting to the ED with syncope or near syncope. Out of 50 variables, they identified 26 predictors that were associated with a serious outcome within 7 days of the index event. They subsequently derived the San Francisco Syncope Rule [history of congestive heart failure, hematocrit<30%, abnormal ECG result (new changes or non-sinus rhythm), complaint of shortness of breath, and systolic blood pressure <90 mmHg during triage] that was prospectively validated in 791 consecutive ED visits with syncope. In the validation study, the San Francisco Syncope Rule classified 52% of the patients as high risk within 30 days, potentially decreasing overall admissions by 7%. If applied only to the 453 patients admitted, the rule could have decreased admissions by 24%. The authors concluded that this rule had a high sensitivity and specificity and is thus valuable in risk stratification. While the San Francisco Syncope Rule reinforced most of the key risk factors identified in the guidelines,26 external validation attempts have failed.2730 In a recent study by Dipaola et al.,31 the San Francisco rule was not superior to clinical judgment with a sensitivity and specificity equal to 81 and 63% vs. 77 and 69%, respectively. Moreover, application of this rule increased admission rate from 34 to 40%. In another Canadian study,32 the application of the rule resulted in even a greater increase in the admission rate from 12 to 69%.

In the recently published ROSE (Risk Stratification of Syncope in the Emergency Department) study, prospective derivation (n= 550) and internal validation (n= 550) of a clinical decision rule were performed at a single centre incorporating B-type natriuretic peptide (BNP) as a major predictor of serious outcome in the first month following the index event.18 In the validation study, the authors found that the ROSE rule had a sensitivity and specificity of 87.2 and 65.5%, respectively, and a negative predictive value of 98.5%. An elevated (BNP) concentration alone was a major predictor of serious cardiovascular outcomes (8 of 22 events, 36%) and all-cause deaths (8 of 9 deaths, 89%). The authors concluded that the ROSE rule and BNP measurement might be valuable risk stratification tools in the ED for patients presenting with syncope. In the same study, the ROSE rule was shown to miss four patients with serious outcomes and to save 149 unnecessary admissions/1000 patients. As of today, external validation of the ROSE rule has not been performed.

In a multi-site Italian study including 676 ED patients with a mean age of 59 years, 41 (6.1%) patients required a major therapeutic procedure or died during the 10-day follow-up period.19 Correlates of the 10-day outcomes included an abnormal electrocardiogram, concurrent trauma, lack of prodromal symptoms, and male gender. Recently, Sun et al.20 developed a 30-day high-risk score in a large cohort (n= 2584) of patients above the age of 60 years presenting with syncope. The score was generated by summing high-risk predictors including age >90 years, male gender, history of an arrhythmia, triage systolic blood pressure >160 mmHg, abnormal electrocardiogram, and abnormal troponin I level. Using this score, the authors were able to stratify patients into low (2.5%), intermediate (6.3%), and high (20%) event rate categories. The estimated overall admission rate in that study was 43%.

In summary, the above tools were primarily aimed at helping the clinician in the ED to identify patients who could be safely discharged. Therefore, by design they all had a very good negative predictive value of the order of 2–3% but a very low specificity. Indeed, these studies showed that only a small percentage of patients admitted suffered a life-threatening event during the few days following the initial presentation. As a consequence, these tools were unable to reduce uneventful admissions compared with current clinical practice and in some cases resulted in an increase in the admission rate.

The Faint-Algorithm

The Faint-Algorithm was designed to help the physician follow the best evidence-based management for each individual patient. Contrary, to risk score-methods, which estimate the risk based on few baseline variables, the Faint-Algorithm is based on a comprehensive structured evaluation of all the important information derived from the initial evaluation in the ED. These data are entered in a web-based interactive algorithm, which allows the physician to easily follow the recommendations of the guidelines. The rationale for using a web-based online interactive decision-making software was to help the physician follow the diagnostic pathway and recommendations of the guidelines during the triage process. It was never intended to replace clinical judgment. In a recent study by Brignole et al.,33 the implementation of the European Society of Cardiology guidelines on syncope reduced the admission rate from 47 to 38%.14 In the present study, when referencing the Faint-Algorithm, 58% of admissions were inappropriate. Moreover, 6% of discharges were also inappropriate. Admittedly, the application of the Faint-Algorithm would have also missed six serious events among discharged patients. However, five of these events were syncopal recurrences without any complications in patients with no cardiac abnormalities. Recent data in the literature show that these events can be managed effectively and safely at home by means of prolonged ECG monitoring instead of hospitalization.34

When evaluating the potential effects of the algorithm on cost, 61 admissions could have been prevented leading to an overall 52% reduction in admission costs to health payers. Among the 61 patients who were inappropriately admitted, a total of 1998 h (83.2 days) were spent in the hospital with an average stay of 33 h (1.4 days). By avoiding the number of inappropriate admission, the algorithm would have saved health-care payers the cost associated with 83.2 days of hospital admission for every 254 patients with syncope, i.e. 327.7 days/1000 patients. Patients with appropriate admissions spent a total of 2832 h (118.0 days) in the hospital with an average stay of 60 h (2.5 days). Assuming that the duration of hospital stay of patients who were inappropriately discharged is similar to the duration of hospital stay of those who were appropriately admitted, by avoiding eight inappropriate discharges, the algorithm would have resulted in an added cost to health care payers corresponding to 20 days for every 254 patients with syncope, i.e. 78.7 days/1000 patients. In total, the application of the algorithm would have saved health care payers the cost associated with 63.2 days of hospital admissions per 254 patients with syncope (83.2 −20), i.e. 248.8 days/1000 patients. In addition, the algorithm could have eliminated the cost related to adverse events due to improper discharges and future malpractice liability.

Limitations

The study was retrospective and the diagnosis of syncope was derived from a billing code instead of defined criteria. A systematic prospective study using well-defined criteria is needed to determine the clinical utility of the proposed approach in the ED. The admission criteria incorporated in our algorithm require echocardiographic data that may not be available to some EDs, thus limiting the application of the software at these sites. Outcome measures for syncope are non-discrete by nature and defined via a consensus of experts. It is, therefore, impossible to determine the sensitivity of any decision set for individual rare serious outcomes without an extremely large cohort. All patients enrolled in this study came from a single centre, although the demographics appear to be similar to those reported by other investigators. Finally, the sample size is relatively small. A multi-centre study is needed to confirm our findings in a larger patient cohort. Despite the above limitations, we believe that the present study points out the shortcoming of the current approaches in the management of patients with syncope and highlights the potential benefits of the Faint-Algorithm.

Conclusion

With the current practice at a major university teaching hospital, the rates of inappropriate discharges and unnecessary admissions were high. While no algorithm should ever replace clinical judgement, our findings highlight the need for more cost-effective approaches to aid physicians when treating patients presenting with syncope. The guideline-based Faint-Algorithm created at the University of Utah provides promise but requires further prospective evaluation.

Conflict of interest: M.H. and M.B. are the co-inventors of the software validated in the proposed study. The University of Utah TCO office recently submitted a provisional patent application. Both physicians have a financial interest in the start-up company that has exclusive rights to the software product.

References

View Abstract